Road Map Generation Using Missing-Feature Guided Re-Recording
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing map data generation methods are costly due to the use of high-end equipment and suffer from incomplete data capture with lower-end cameras, leading to missing crucial features for accurate map generation.
Innovation Solution
A system that uses lower-cost cameras to collect 2D training images, constructs 3D maps, identifies potential missing features, and selectively targets areas for re-recording or high-end equipment to ensure comprehensive coverage, integrating neural networks for feature likelihood determination.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If traditional expensive specialized equipment with elevated cameras is used, then map data completeness is improved, but cost increases significantly
Solution Approach 1:
The patent replaces expensive specialized equipment with lower-cost cameras that can be deployed on multiple vehicles. Instead of using a few high-end cameras, the system uses many inexpensive cameras to achieve the same data collection goal, reducing the cost per kilometer significantly while maintaining map data completeness through aggregated data from multiple sources.
Solution Approach 2:
The patent combines data from multiple low-cost cameras mounted on different vehicles to achieve the same mapping quality that would require a single expensive elevated camera system. By merging images from multiple vehicles passing through the same area, the system recovers complete map features that any single low-angle camera might miss.
2Ease of manufacture
If lower-end cameras are used to reduce cost, then cost decreases, but map feature detection accuracy deteriorates due to missing features
Solution Approach 1:
The patent implements a feedback mechanism where the system identifies map features that are missing or incomplete from initial image processing, then selectively triggers re-recording missions in those specific areas. This feedback loop ensures that critical map features are captured accurately by sending vehicles back to re-image areas where features were missed, thereby improving detection accuracy without requiring all areas to be re-recorded.
Solution Approach 2:
The patent performs preliminary identification of missing map features before triggering re-recording missions. By analyzing initial images to detect which map features are absent or incomplete, the system can pre-plan targeted re-recording routes to specifically capture the missing features, rather than re-recording entire areas.
3Ease of manufacture
If the same road is covered multiple times with lower-cost cameras, then cost per km decreases, but time consumption increases
Solution Approach 1:
The patent applies local quality by differentiating between areas that require re-recording and areas that have sufficient data. Instead of uniformly re-recording all areas multiple times, the system identifies specific local regions where map features are missing and targets only those areas for re-recording, thereby reducing overall time consumption while maintaining cost efficiency.
Solution Approach 2:
The patent performs preliminary analysis to identify which specific road segments and features require re-recording before dispatching vehicles. This pre-planning allows the system to optimize re-recording routes and avoid unnecessary revisits to areas that already have sufficient map data, reducing time consumption.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A system configured for managing orders is disclosed. The system may include one or more processor(s) which may collect first 2D training image data; collect second 2D training image data; construct a 3D map for the geographical area based on the first training image data and the second training image data; determine a likelihood of a potential missing feature in the 3D map based on the first 2D training image data and the second 2D training image data; collect third 2D training image data comprising third map images of the geographical area acquired by the one or more image acquisition apparatus if the likelihood of the potential missing feature is above a predetermined threshold; and generate the road map based on the first 2D training image data, the second 2D training image data and the third 2D training image data.